The association between patch metrics and roosting site (n = 31) suitability of Lyle’s flying fox (Pteropus lylei) in 26 Central Eastern and Western provinces of Thailand was quantified. Land use classes with 90-m resolution were identified based on various vegetation and land cover types to calculate patch metrics using FRAGSTATS. Then, Maximum Entropy Modeling (MaxEnt) was performed using patch metrics covariates to produce a predictive potential distribution map. The results indicated that patch contiguity (contiguity index, 63.7%), patch area (29.3%), and patch shape complexity (shape index, 5.7%) are the most influential patch metrics, all of which have negative effects on roosting site suitability. In total, 13,222 small patches were considered highly suitable patches, with a mean area of 0.921 ± 0.698 (SD) ha, which accounted for 122,090 ha (2.04%) of the study area. Roosting sites predicted from the model were consistently associated with occurrences of roosting sites observed in temples; such habitats likely provide shelter from external threats for colonies roosting in a human-dominated landscape.
